一种半监督的彩色图像分割方法

提出一种基于半监督EM聚类的彩色图像分割方法, 算法利用了有限的人工信息, 即在图像上点击有限的几个点以标识对应区域之间的关系, 从而得到满足给定限制的精确图像分割结果。算法首先对图像进行量化处理, 而后在量化后的色彩空间中集成先验的分割信息进行色彩聚类。实验表明算法运行速度快, 分割效果好, 具有很高的应用价值。     

基于模糊连接度的FCM分割方法在医学图像分析中的应用

图像分割的一个重要应用领域是医学图像的分割.我们针对医学图像的模糊特点和实际应用的要求,结合模糊连接度阈值分割和模糊C均值聚类两种分割方法的优点,提出一种新的交互式医学图像分割方法.首先计算整幅图像的模糊连接度,通过闽值分割提取出感兴趣的对象,并将模糊连接度作为图像的冗余特征;然后在由冗余特征和原图像特征构成的二维聚类空间中,利用模糊C-均值聚类方法优化上一步骤的分割结果,提高分割准确度.我们以CT和MR图像为实验对象进行了验证,实验结果表明这是一个有效的方法.     

基于两幅视图的纵向插补方法

目前,纵向移动视点的多视图插补仍然是高质量漫游系统的一个难点问题。为大家熟知的进入式漫游(TIP)由于采用单幅图像放大的方法造成图像分辨率下降,而多幅图像分别构造TIP模型的方法在漫游过程中存在着明显跳跃的不足。基于两幅视图的纵向插补方法在获得摄像机中心处于同一条直线上的两幅图像后,用平面射影变换求对应的射影区域;然后利用这两幅图像的焦距关系,计算边缘点的空间位置;最后利用边缘点确定射影区域的空间位置和像素RGB值。实验证明该算法简单,容易实现,漫游效果平滑。     

皮肤黑素细胞肿瘤图像综合分割方法研究

目的 本文针对黑素细胞肿瘤（Melanocytic Tumor MT）图像情况复杂，较难分割的问题，提出了一种综合数字图像分割算法，探讨MT的早期诊断。方法 首先应用统计区域融合方法（SRM）实现图像分割成多块纹理一致的区域。然后对图像以HSV彩色空间的H和S分量为特征，使用K均值聚类算法将图像聚为9类。最后，将聚类结果在HSV彩色空间的H和S分量值分别映射到[0，1]区间，再分别对H分量和S分量取阈值，得到最终的边界分割结果。结果 对MT图像能够按照其纹理差异将其有效划分为多个区域，较为准确标识出皮损区域。结论 综合对多种方法结果的对比，本方法优于传统的大津阈值法、K均值法和活动轮廓法。同时对过去基于SRM的MT图像分割方法进行了改进，在处理复杂MT图像时效果明显好于传统方法。     

基于点云辐射场的新视角图像生成网络

新视角图像生成是指使用同一场景、多个位置和角度拍摄的一组图像，生成新颖视角位置处的场景图像，在真实拍摄的城市建筑场景图像中存在移动的瞬态对象，对静态建筑造成遮挡，导致生成的新视角图像模糊、存在伪影甚至建筑物生成错误。针对于此，本文提出了一种基于点云辐射场的新视角图像生成方法。首先，通过瞬态对象过滤模块得到静态建筑图像；然后，分别经过2D CNN和3D CNN网络获取图像特征和图像点云信息，二者结合构建神经点云；最后，输入多层感知机网络获取场景采样点位置的体积密度和颜色信息，并指导体渲染模块生成新视角位置的图像。在Photo Tourism数据集与最新的几种方法进行比较，实验结果和可视化结果表明，所提方法在数据集上有较好的效果表现，有效地解决了生成图像中存在伪影和建筑物生成错误的问题。     

一种综合的统计纹理分割方法

本文提出了一种无需任何先验知识的纹理图像分割系统。该系统采用模板滤波来提取纹理特征值，并用动态聚类法估计各类纹理的统计参数；然后引入空间约束条件对初始概率分布函数进行修正，并根据修正后的概率分布函数来确定各个象素所属的纹理类别。实验证明该系统无论是对人工纹理还是自然景物，都能取得良好的划分效果     

基于色度学准则分析的免疫组化彩色图像C-均值聚类分割技术研究

免疫组化彩色图像分割在免疫组化染色定量分析中有重要的应用价值。C-均值聚类算法(CMA)是根据一定的相似性准则将图像分成C类。由于运算十分耗时,直接限制了C-均值聚类算法在彩色图像分割领域的应用。本文针对免疫组化彩色图像特点,提出了分割免疫组化彩色图像的色度学准则,即:用每个像素的R分量减去B分量,根据其差值是否大于0将相应的像素分为两大类:(R-B)≥0类和(R-B)<0类,自动分离出图像的阳性细胞区域和阴性细胞区域。在此基础上我们改进了C-均值聚类分割的方法:①针对上述两大类图像的色彩分布的特点确定初始聚类中心;②分别对上述两大类像素在一个色度学空间聚类;③根据每次迭代过程中聚类中心的变化趋势来预测下一次迭代后可能的聚类中心,从而减少迭代的次数。改进之后的C-均值聚类分割方法减少了聚类的样本数量,降低了算法的复杂度,且由于迭代次数的减少,运行速度得到了提高。实验结果表明,本文建立的技术方法能有效地分割免疫组化彩色图像。     

SAM法在多光谱图像分类中的应用

不同类型的地物具有不同的反射光谱, 在多维光谱空间中构成不同的特征向量, 这便是我们利用多光谱图像区分地物的物理依据。多光谱图像分类的精度受到多种因素的影响, 如模式样本集的数据结构、分类特征的提取、距离测度、分类准则、分类数目等。本文引入光谱匹配识别中的典型方法——光谱角度匹配法(SAM法), 通过与欧氏距离的比较分析与综合, 提出一种可以兼顾到光谱亮度与光谱向量方向(近似为光谱形状) 的距离测度, 并应用于K-均值动态聚类中, 通过与传统分类方法的比较, 证明这种方法的有效性。     

实景树木参数化建模方法研究

受限于航空倾斜摄影图像的拍摄角度，三维重建场景中的树木模型存在大量失真、变形和空洞。而更为精细的树木建模方法，如基于近景图像、激光点云的树木建模方法效率低且成本较高。因此，提出一种具有真实感的实景树木参数化建模方法，对真实场景中的航空倾斜摄影图像和树木三维点云进行特征提取，归纳各类树木的几何外形、拓扑结构和纹理材质，设计一套适用于各类树木的生长规则，最终实现实景树木多细节层次建模。实验结果表明，该方法能够保证模型整体的完整性和细节的真实性，同时满足场景渲染要求，为三维场景重建、数字林业和智慧城市等应用提供有效的技术支撑。     

基于MeanShift和Harris算子圆网印花机对花检测的研究与设计

为了实现基于智能相机的圆网印花自动对花系统,本文提出了基于均值漂移(MeanShift)算法的彩色印花图案分割技术和基于Harris角点检测的块匹配算法。将扩展形式的均值漂移算法用于印花图案的分割,实验证明该算法能对印花图像进行良好的分割。把已分割标准图像的各套色区域提取出来,采用Harris算子进行特征点检测,并以这些特征点为中心,选定标准匹配块,在处理好的实时采集的图像中找到最佳匹配块,计算对花误差。通过仿真和实验验证算法是可行性的,为实现在线对花检测闭环控制打下基础。     

Respiratory-driven lung tumor motion is independent of tumor size, tumor location, and pulmonary function

PURPOSE: To determine whether superior-inferior lung tumor motion is predictable by tumor size or location, or pulmonary function test results. METHODS AND MATERIALS: Superior-inferior tumor motion was measured on orthogonal radiographs taken during simulation of 22 patients with inoperable lung cancer diagnosed by orthogonal radiographs. RESULTS: The tumor size averaged 5.5 +/- 3.1 cm (range 1.5-12 cm). Seven of 11 central tumors demonstrated some motion compared with 5 of 11 peripheral tumors. Four of 5 upper lobe tumors moved compared with 8 of 17 tumors that were either middle or lower lobe lesions. The mean fourth rib motion was 7.3 +/- 3.2 mm (range 2-15). The mean FeV(1) was 1.8 +/- 1.2 (range 0.55-5.33. The mean diffusing capacity of the lung for carbon monoxide was 14.0 +/- 6.5 (range 7.8-21.9). The mean total lung capacity was 6.5 +/- 1.2 (range 3.3-8.4). None of these parameters correlated with tumor motion. Although lateral tumor motion could not be consistently determined, 1 tumor moved 10 mm anterior-posteriorly. CONCLUSIONS: Lung tumors often move significantly during respiration. Tumor motion is not predictable by tumor size or location, or pulmonary function test results. Therefore, tumor motion must be measured in all patients. Measurement in three dimensions will likely be necessary to maximize the irradiated lung volumes or choose beam arrangements parallel to the major axis of motion.     

基于高噪声深度图像的树木模型重建

对现实世界中的植物模型进行真实准确的重建一直是计算机图形学的一个重要目标。基于激光扫描仪提供的深度图像的树木重建近年得到越来越多的关注。当扫描距离过远或者扫描精度较低的情况下深度图像带有较大的噪声,这给重建带来了很大的困难。本文提出了一种针对高噪声深度图像的树木自动重建技术,本算法以主曲率分析为基础,首先通过二次曲面拟合计算各个点对应的主曲率方向,之后利用各个点云与周围点云的主曲率相似度将噪声点云去除,利用点云分割技术将属于不同树枝的点云初步分离,最后从分割数据中提取树枝的骨架点以及半径,完成重建。利用本方法 ,我们对一个具有高噪声的深度图像进行了处理,从中提取出了树木的三维模型。本算法能够利用之前算法无法处理的高噪声的深度图像重建树木模型,降低了树木重建算法对输入数据精度的要求,同时减少了手工交互。     

On the incorporation of multi-modality image registration into the radiotherapy treatment planning process

A technique is presented that allows the direct use of physiological image sets in the radiation therapy treatment planning process. When fused to the treatment planning CT, physiological image studies may allow one to define physiological tumor subvolumes consisting of areas of possible chronic hypoxia, areas of high perfusion, areas of high diffusion, and areas containing high choline concentrations. These physiological tumor subvolumes could be selectively boosted to increase local control of malignant brain tumors once one has determined which of these physiological tumor subvolumes predicts for local tumor recurrence after conventional radiotherapy. In this technique a user assisted automatic registration technique is used that is based on an analytical estimate for the transformation matrix needed to register two rigid bodies. The only user input needed is three non-collinear points selected based on landmarks in the primary image and the corresponding three points in the secondary image. Since this registration technique uses two sets of at least three user-defined landmark points each of which has some selection error associated with it, the final registration will have an error that depends only on the selection error associated with the point sets. Since physiological image studies are acquired at the same setting as the T1- w MRI their spatial orientation with respect to the T1- w MRI is known. Therefore, the registration of multiple physiological image studies to the treatment planning CT can be accomplished by first correlating them to the T1- w MRI, and in a second step the T1- w MRI is then registered to the treatment planning CT. The desired registration of the physiological image studies to the treatment planning CT is then accomplished by simply composing the appropriate transformation matrices.     

一种新的用于图像稳定的特征点检测方法

二维特征点的检测和提取是进行图像配准、目标识别和运动匹配的关键技术。针对不同的后继任务, 检测和提取的方法也有所不同。在多目标跟踪和识别技术中, 图像稳定是必要的处理过程, 而如何找到对应的特征点则是其难点所在。本文基于Gabor小波变换, 提出了一种新的分散型自适应策略。该方法能够迅速、有效地在前后两帧图像中找出可能的对应特征点, 以便进行特征点匹配, 从而完成图像稳定。实验表明, 本文方法检测到的特征点能够确切代表两帧图像间的运动情况, 从而为图像稳定提供了可靠的基础。     

Value of multi-planar CT images in interactive dosimetry planning of intracavitary therapy

A method of intracavitary treatment planning and dosimetry analysis which uses multi-planar reconstructed computerized tomography (CT) images is presented. The aim of the method is to improve ability to precisely locate clinical reference points, to fully define pertinent anatomic structures and to provide dose distributions and their relationship to these structures in multiple planes. Our approach is based on interactive treatment planning and point dose display on sagittal and coronal reconstructed CT images as well as the usual transaxial image. The advantages of clinical evaluation of isodoses directly on multi-planar CT images are assessed. These include precise anatomic and dose relationships between the cervix and paracervical structures, the bladder, rectum and pelvic node-bearing sites. Problems of image magnification, blurred images and inadequate resolution attendant to orthogonal radiographs, which are the basis of current techniques, are minimal. Analysis and results of the method and a comparison with the technique of orthogonal radiographs are presented for a demonstration case.     

A mathematical model for cisplatin cellular pharmacodynamics

A simple theoretical model for the cellular pharmacodynamics of cisplatin is presented. The model, which takes into account the kinetics of cisplatin uptake by cells and the intracellular binding of the drug, can be used to predict the dependence of survival (relative to controls) on the time course of extracellular exposure. Cellular pharmacokinetic parameters are derived from uptake data for human ovarian and head and neck cancer cell lines. Survival relative to controls is assumed to depend on the peak concentration of DNA-bound intracellular platinum. Model predictions agree well with published data on cisplatin cytotoxicity for three different cancer cell lines, over a wide range of exposure times. In comparison with previously published mathematical models for anticancer drug pharmacodynamics, the present model provides a better fit to experimental data sets including long exposure times (approximately 100 hours). The model provides a possible explanation for the fact that cell kill correlates well with area under the extracellular concentration-time curve in some data sets, but not in others. The model may be useful for optimizing delivery schedules and for the dosing of cisplatin for cancer therapy.     

放射治疗射野片的拍摄和分析

目的　讨论如何实施和完善放射治疗射野片的拍摄和分析。方法　采用双曝光技术进行放射治疗射野片的拍摄 ;在首次曝光时加入射野参考坐标标记点 ;以铅点影像所勾画成的十字线为实际射野的参考坐标 ,与模拟机的计划摄片进行比较分析。结果　射野片上铅点和挡块轮廓的影像都较为清晰 ,但射野内组织影像的分辨率和图像对比度逊于常规的模拟机摄片。不过 ,仍可以从射野片上分辨出照射野内各组织的结构和位置 ,从而分析和判断实际照射野是否达到放射治疗计划的要求。结论　射野片的实时性不如电子射野影像系统 ,但其具有成本低廉、容易实施的特点 ,可成为放射治疗执行阶段中重要的质量控制手段 ;参考标记点的设置是射野片分析工作中不容忽视的环节。     

Clinical Feasibility of Multiplanar Reconstruction Images of Temporal Bone CT in the Diagnosis of Temporal Bone Fracture with Otic-Capsule-Sparing Facial Nerve Paralysis

The aim of this study is to evaluate the feasibility of multiplanar reconstructive (MPR) imaging of temporal bone CT in the diagnosis of temporal bone fracture with oticcapsule-sparing facial nerve paralysis. Twelve patients with traumatic facial nerve paralysis with otic-capsule sparing and temporal bone fractures were selected. Multiplanar reconstruction images were obtained with the V-works 4.0 software program (Cybermed, Seoul, Korea) using axial scanning of high-resolution temporal bone CT of the fracture line. The clinical profiles of the patients displaying temporal bone fractures were examined in relation to the findings. Multiplanar images of the fracture line provided information regarding the direction of the external force that fractured the temporal bone. The fracture line was more continuous in the MPR images than in the axial view. All patients showed an imaginary extended fracture line directed toward the otic capsule. The direction of the fracture line toward the middle ear cavity is important, as it may suggest insult to the otic capsule. The MPR image parallel to the fracture line of the temporal bone provides a guideline for the vector of the force that induced the fracture. Thorough investigation of the critical organs during surgical exploration is recommended if the direction of the fracture in the MPR image points toward the otic capsule in the middle ear even if the fracture line relative to the otic capsule is not well defined in the axial or CT view.     

Bildqualitätseinbußen bei der Bildausgabe eines digitalen Röntgenaufnahmesystems

Not all digital radiography systems allow an unrestricted image postprocessing for the radiologist. Depending on the X-ray system there are restrictions in varying brightness and contrast of the image. We compared grey level information that are given to the radiologist by various digital radiography systems. Histograms of differently postprocessed X-ray images were calculated, which were acquired by three different X-ray systems. We found that every modality has its own way of saving and converting the images into DICOM data. The processed image is either sent into PACS with the initial grey level range (10 or 12 bit) or the image is sent with a reduced grey level range due to windowing. The grey level range is associated with the ability of unrestricted windowing in PACS. So, taking an image with a reduced grey level range, the radiologist will face limited potential to optimise the window setting individually afterwards. The loss of image quality due to image transfer from the modality to PACS can lead to an information loss in the diagnostic relevant range.